Thread by @IHME_UW

Thank your feedback, questions, and critiques on our #COVID19 model. We’ve updated our FAQ, and are working to incorporate more feedback into our model and affiliated resources, including our Estimation Updates blog.🧵The following is a thread addressing recently asked questions:

What does our model say about expected new #COVID19 cases?

We are forecasting surges of new cases in Kansas, Missouri, New Jersey, Virginia, and West Virginia in the next week or two.

Where do we expect fall #COVID19 death surges?

We currently projects surges in deaths on Jan 1 (if 95% public masking is not adhered) in these states: AL, AZ, AK, CA, DE, CO, KS, IN, IA, MD, MT, NE, NM, NV, NY, NH, NC, OH, OK, OR, PA, RI, SD, TN, UT, VA, WA.

Has mask use in the United States increased?

According to @premisedata, we saw mask use rise in the United States between April and mid-July, but start to decline in late-July/early-August.

Details and downloadable images here: healthdata.org/acting-data/ma…

Has mask use made a difference?

Yes. Mask use rose in the US (per @premisedata) over June and July at exactly the same time R0 fell. Mask usage sits at 45% in the US compared to 60% globally.

View mask use in our viz tool:
covid19.healthdata.org/united-states-…

Why focus on masks?😷

Governments around the world show no appetite for re-imposing sweeping lockdowns because of its economic impacts. We believe a focus on universal masking will help save lives and the economy.

Where does your seasonality assumption come from?

So far, in our research, COVID-19 transmission correlates strongly with pneumonia seasonality.

What is the predictive validity of our #COVID19 forecasts? AKA: How do our forecasts compare to actuality?

We have 12 week median absolute percent errors from models in June, via @JosephRFriedman:

https://twitter.com/JosephRFriedman/status/1304486880298192896?s=20

Why did we extend forecasts out to January 1st?

We have been forecasting four months out to give governments, policymakers, and healthcare administrators sufficient time to plan for the pandemic. They have asked us to share projections at least 4 months out.

Are you forecasting certainty?

No! You can observe our uncertainty intervals by selecting that option within each chart in our viz tool. We strongly recommend policymakers draw from a variety of models for planning purposes.

What are important assumptions of your model?

We’re going to answer this in four parts.

Important Assumption 1⃣: COVID-19 follows seasonality patterns similar to pneumonia. We've seen clear indications of seasonality from South American countries such as Chile and Argentina that have just emerged from their winter.

Important Assumption 2⃣: The “Current projection” scenario assumes that social distancing mandates will continue to be lifted, but will be re-imposed for six weeks if daily death rates reach 8 per million.

Important Assumption 3⃣: The “Mandates easing” scenario assumes that social distancing mandates will continue to be lifted and will not be re-imposed.

Important Assumption 4⃣: The “Universal masks” scenario assumes that mask wearing will reach 95% in 7 days, and social distancing mandates will continue to ease, but will be re-imposed for six weeks if daily death rates reach 8 per million.

To view/interact with the scenarios and assumptions we’ve just detailed, visit our #COVID19 visualization tool: covid19.healthdata.org/projections

Why do our #COVID19 forecasts show significant digits?

We report the projections our model produces with as much precision as possible and without interpretation. Rounding numbers can be helpful, which we do ourselves in press releases, newsletters, and social media posts.

Why aren’t the model’s details and assumptions more readily available?

We’re going to answer this in three parts.

1⃣ The current version of the model is described here: healthdata.org/research-artic…

2⃣ We are working very quickly to produce projections that can inform the policy discussion in a timely fashion, as requested by many key stakeholders; updates to the methodology are cataloged here: healthdata.org/covid/updates/…

3⃣ All of our COVID-related publications are available here: healthdata.org/covid/publicat…

We welcome continued feedback, comments, questions and critique. Thank you!

@threadreaderapp unroll please